Penguins selectivity for krill size is shown as highly sensitive to the statistical assumptions made in the analysis of data. Nested designs are proposed as the correct approach because lack of independence between krill found in the stomach of a given penguin prevents the pooling of krill lengths from different samples.
Samples taken at Bahía Esperanza are used to illustrate the different approaches. A highly significant linear regression between krill size and time was found.
Full use of the information contained in stomachal contents requires the use of block sampling designs. Replicating samples of the same penguin or the study of length distributions of the various (according to digestion degree) fractions of stomachal contents are proposed as alternative methods.
